LiteRT-LM v0.14.0 migration — litertlm 1.2.0, embeddings 1.0.3 (native per-session sampler + #214 fix)#387
Merged
Merged
Conversation
…que sampler API (Phase 2)
…T v0.14.0 signature (adds LiteRtEnvironment)
… split
v0.14.0 split Dawn out of libLiteRtWebGpuAccelerator into a separate
libwebgpu_dawn.{so,dll} (0.13.1 static-linked it), so the accelerator now
hard-depends on it and desktop GPU fails to load (all backends) without it.
- hooks: add 'webgpu_dawn' to companions (+ 'libwebgpu_dawn' to
windowsExtraLibs for the PE import); bump native version to 0.14.0
- Linux CI: patchelf --set-rpath '$ORIGIN' so each lib finds its
flat-layout siblings at dlopen time
- Windows CI: bundle libwebgpu_dawn.dll in the Collect companion loop
Validated on Linux (Tesla T4): litertlm smoke 24/24 pass.
…ler + #214 fix (litertlm 1.2.0, embeddings 1.0.3) - FFI on v0.14.0 opaque session-config/sampler API; native-v0.14.0 (7 platforms). - #214 GPU output-garbage fixed; macOS/Linux GPU + Windows CPU/NPU verified. - Web @litert-lm/core 0.12.1 -> 0.14.0 (text path). - Add community models (SmolLM3-3B, Phi-4-mini-reasoning, Qwen2-VL, SmolVLM2, LLaVA-OneVision) to example + README. - macOS: re-sign companion frameworks after install_name_tool (fixes CODESIGNING Invalid Page). - Known: Windows discrete GPU regressed upstream (LiteRT-LM #2957) — use CPU/NPU.
- installation.md: bump the missed 4th web @litert-lm/core snippet 0.12.1 -> 0.14.0. - CLAUDE.md: native tag native-v0.12.0-a/0.13.1-a -> native-v0.14.0. - README/MIGRATION/DESKTOP_SUPPORT: version pins -> 1.3.2/1.2.0/1.0.3. - README: Qwen2-VL/SmolVLM2/LLaVA run everywhere .litertlm does (Mobile ✅); FastVLM stays desktop-only. - litertlm CHANGELOG: narrow #214 claim to verified-on-Xclipse. - internvl35 smoke header: 500M -> 1B.
… docs (PR #387 review) - Windows discrete-GPU regression now flagged on every desktop surface (README, DESKTOP_SUPPORT, website desktop/troubleshooting, litertlm README) with a consistent note + LiteRT-LM #2957; CPU/NPU guidance. - README "What's new in 1.3": add LiteRT-LM v0.14.0 runtime (native per-session sampler + #214) and the 5 new community models. - DESKTOP_SUPPORT/desktop.md: native tag -> native-v0.14.0; replace obsolete downstream-sampler-patch narrative with native per-session sampler; add libwebgpu_dawn to bundle lists; disambiguate "0.14.0" (plugin vs LiteRT-LM); macOS symlinks -> .framework wording. - Website models/multimodal/thinking-mode/function-calling: add the 5 new models to match README. - jaspr build verified (no SSG errors).
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
LiteRT-LM v0.14.0 migration
Moves the native inference stack to LiteRT-LM v0.14.0 and ships the follow-on package releases.
Releases
flutter_gemma_litertlm1.1.0 → 1.2.0flutter_gemma_embeddings1.0.2 → 1.0.3flutter_gemma1.3.1 → 1.3.2 (docs: README model list + web setup snippet)What's in it
@litert-lm/core0.12.1 → 0.14.0 (text path; API-compatible — verified against the TS declarations).install_name_tool(was tripping CODESIGNING "Invalid Page" at dlopen).libwebgpu_dawn.{so,dll}.Validation
native-v0.14.03-way checksum consistency verified across all 7 platforms; fresh macOS build fetches + links against the release.flutter build web).Windows discrete GPU (WebGPU/Dawn, D3D12) crashes during
engine_createin the upstream prebuiltlibLiteRtWebGpuAccelerator.dll— a regression from v0.13.1, isolated to upstream (unmodified prebuilt; CPU/NPU work with the same host lib; DX runtime byte-identical to v0.13.1). Filed upstream: google-ai-edge/LiteRT-LM#2957. Use CPU or NPU on Windows. macOS/Linux GPU and Windows CPU/NPU are unaffected.